Domain.com.au Property Scraper (Australia)

Scrapes houses and apartments for rent or sale from Domain.com.au — Australia's #1 property portal (larger than realestate.com.au).

Public data only. No login, no cookies, no browser.

The one thing you need to know before using this

The search page is behind Akamai Bot Manager, which serves a disguised HTTP 200 "behavioral content" challenge page to most TLS fingerprints — it looks like a normal response by status code alone. Of every profile tried this session (chrome124, chrome120, edge101, firefox133, safari17_0, and three primp profiles), only chrome99_android returned genuine content. This actor pins to that profile as the pool leader rather than using the portfolio's usual default, and checks response bodies for challenge markers before trusting a 200 (same defensive pattern as every other actor here, just against a target where it actually triggers).

What you get

Two record types share one dataset, told apart by recordType.

PROPERTY — one row per listing

Search rows (listing) already carry everything: price, full street address with coordinates, bedrooms, bathrooms, property type and every photo. No separate detail fetch exists in this actor — none was found to be necessary this session (search already has it all, same architecture decision as bienici-properties-scraper elsewhere in this portfolio).

SEARCH_SUMMARY — one row per location searched

Domain's own totalListings/totalPages plus how many pages this run actually fetched.

Input

Field	What it does
Rent / sale	which side of the market (note: the real path segment for buying is sale, not buy)
Locations	Domain slugs: melbourne-vic-3000, bondi-nsw-2026…
Max properties / max pages per location	pagination caps

Example

{
  "transactionType": "rent",
  "locations": ["melbourne-vic-3000", "bondi-nsw-2026"],
  "maxItems": 100
}

Notes on reliability

• Active WAF (Akamai) — see above. Residential proxy is the default on cloud runs on top of the profile pin; a fresh IP does not fix a bad fingerprint, but a bad IP can still add a second failure mode.
• Unknown locations fail cleanly as not_found — a clean HTTP 404, never a silently-wrong nationwide baseline.
• A page past the end is a clean HTTP 400 (page=999 on a 50-page search, verified live), reported as past_page_range — not a silent empty/wrong page.
• buy 404s — the correct path segment is sale; confirmed live so the actor doesn't offer the wrong one.

Output envelope

Every record carries _input, _source and _scrapedAt. Upstream field names pass through verbatim under listing — no renaming. _source is S1-nextdata-search on every row.

See CRAWLING_METHOD.md for the full reverse-engineering trail, including the complete TLS profile ladder results.